Abstract: An electronic device including a housing, a display panel with a flexible substrate, and a flexible printed circuit, in which the housing comprises a flat surface and a side surface comprising a curved region, the display panel comprises a first region that overlaps with the flat surface and a second region that overlaps with the side surface, the display panel comprises a convex portion that comprises a region overlapping with a flexible printed circuit, and the display panel comprises a display portion in the first region and the second region.

Abstract: A tiled electronic device includes a plurality of display panels, and at least one of the display panels includes a flexible substrate, a pixel, and two signal wires. The flexible substrate has a display portion and a bent portion connected to the display portion. The pixel is disposed on the display portion. The signal wires are disposed on the flexible substrate, and electrically connected to the pixel. Each of the signal wires has a first segment disposed on the display portion, and a second segment disposed on the bent portion. The two first sections have a first pitch, and the two second sections have a second pitch. The first pitch is different than the second pitch.

Abstract: Methods and systems are disclosed for creating a shared augmented reality (AR) session. The methods and systems perform operations comprising: receiving, by a client device, input that selects a shared augmented reality (AR) experience from a plurality of shared AR experiences; in response to receiving the input, determining one or more resources associated with the selected shared AR experience; determining, by the client device, that two or more users are located within a threshold proximity of the client device; and activating the selected shared AR experience in response to determining that the two or more users are located within the threshold proximity of the client device.

Abstract: A content transmission method includes obtaining, by a first terminal device, a first instruction instructing the first terminal device to perform content transmission with a second terminal device sending, by the first terminal device, to the second terminal device, and in response to the first instruction, first information instructing the second terminal device to determine that that the first terminal device transmits content to the second terminal device. Receiving, by the first terminal device, content transmission direction indication information from the second terminal device, determining, by the first terminal device, first content, and sending, by the first terminal device, the first content to the second terminal device to display the first content.

Abstract: An electronic device is provided. The electronic device includes receiving display information of the target device from the target device using a communication module, expanding or reducing a flexible display of the electronic device based on the display information of the target device, generating mirroring data including a content area in which content is displayed and an additional information area in which additional information is displayed based on the display information of the target device, and transmitting the mirroring data to the target device using the communication circuit.

Abstract: A display control method includes causing a display device to display a processing image in which a first image representing a note corresponding to a synthesized sound and a second image representing a sound effect are arranged in an area, in which a pitch axis and a time axis are set, in accordance with synthesis data that specify the synthesized sound generated by sound synthesis and the sound effect added to the synthesized sound.

Abstract: Embodiments provide a method and system for audio generation from contextual text input is provided. The disclosure gives due importance to the granularity of the content. The system allows the user to specify the properties of the audio to be generated. Here, context is used to identify the importance of a particular sound over the others and thus automatic adjustments of the audio output to give a more realistic feel. The system generates dataset for training audio models. The user can give input query in natural language and the audio requested will be generated for training and developing the necessary classification or other necessary audio models. The system provides a feature of automated fine-tuning of the model parameters to suit the new automatically collected training data. Furthermore, the system provides a pre-trained inbuilt model repository with audio models belonging to the main categories of noises.

Abstract: An example playback device is configured to play back a first media item and, while playing back the first media item, receive a voice command to enter a preview mode in which the playback device plays back one or more second media items for a predetermined duration that is shorter than a total duration of the one or more second media items. Responsive to receiving the voice command, the playback device enters the preview mode, thereby (i) suspending playback of the first media item and (ii) initiating playback of the one or more second media items for the predetermined duration. The playback device subsequently terminates the preview mode, thereby (i) terminating playback of the one or more second media items, and (ii) resuming playback of the first media item.

Abstract: A covert sports communication system has a transmitter configured to wirelessly transmit coded audio track selection signals. The system includes a receiver that may be worn within headgear. The receiver includes stored audio tracks and is configured to receive coded audio track selection signals from the transmitter, decode the audio track selection signals, and play at least one stored audio track in accordance with the decoded audio track selection signals. The audio tracks may include audio instructions to a participant of a sporting event.

Abstract: Example techniques involve invoking voice assistance for a media playback system. In some embodiments, a NMD stores in memory a set of command information comprising a listing of playback commands and associated command criteria. The NMD captures a voice input and detects inclusion, within the voice input, of one or more particular playback commands from among the playback commands in the listing. In response, the NMD selects a local voice assistant that supports (a) one or more additional playback commands relative to a cloud-based VAS and (b) fewer non-playback commands relative to the cloud-based VAS, determines, via the local voice assistant, an intent in the captured voice input, and performs a response to the determined intent. The NMD foregoes selection of the cloud-based VAS when the local voice assistant is selected.

Abstract: In response to a user interacting with a tangible peripheral assistant control device (e.g., depressing a button of the device), causing an automated assistant to perform one or more actions. The action(s) performed can be based on input previously provided by the user in configuring the peripheral assistant control device. The action(s) performed in response to interaction with the peripheral assistant control device can vary based on one or more conditions, such as which user is currently active, where the peripheral assistant control device is currently located (which can optionally be inferred based on which of multiple assistant computing devices the button is paired with), and/or the current state of one or more smart devices and/or other devices (e.g., as determined based on a device topology). A utility of the peripheral assistant control device can be automatically extended beyond what was specifically requested by a user during configuration.

Abstract: Techniques for routing a user command to a speechlet and resolving conflicts between potential speechlets are described. A system determines an intent of an input command. The system also receives context information associated with the input command. The system determines speechlets (e.g., speechlets and/or skills) that may execute with respect to the input command given the intent and the context data. The system then determines whether conditions of routing rules, associated with the speechlets, are satisfied given the context data. If the conditions of only one routing rule are satisfied, the system causes the speechlet associated with the routing rule to execute with respect to the input command. If the conditions of more than one routing rule are satisfied, the system may determine a speechlet to execute with respect to the input command based on the speechlets' priorities in a list of speechlets and/or based on potential output data provided by the speechlets.

Abstract: A method includes determining, by an assistant executing at one or more processors, a default group of actions that the assistant is configured to execute in response to receiving a particular audible command. The method includes determining, by the assistant, based on the default group of actions and a user profile associated with a particular user, a custom group of actions that the assistant is configured to execute in response to receiving the particular audible command from the particular user. The method also includes receiving, by the assistant, an indication of the particular audible command, and determining, by the assistant, whether the indication of particular audible command originated from the particular user. The method further includes, responsive to determining that the indication of particular audible command originated from the particular user, executing, by the assistant, each action from the custom group of actions.

Abstract: Disclosed is a parametric feature-based 3D CAD system that allows multiple users to simultaneously edit a parametric feature-based 3D CAD model consisting of 3D parts and assemblies of those parts (3D Model). Several CAD users, each using their own computer, phone, or tablet, can edit the same 3D Model at the same time. Editing may be separate and simultaneous—there is no need for users to worry about locking, checking out, or otherwise restricting each other's access to 3D Models. Within each user's editing interface, the model may be presented in different representations, with edits to any representation applied and displayed across all representations.

Abstract: A computer-implemented method of machine-learning including obtaining a dataset of 3D point clouds. Each 3D point cloud includes at least one object. Each 3D point cloud is equipped with a specification of one or more graphical user-interactions each representing a respective selection operation of a same object in the 3D point cloud. The method further includes teaching, based on the dataset, a neural network configured for segmenting an input 3D point cloud including an object. The segmenting is based on the input 3D point cloud and on a specification of one or more input graphical user-interactions each representing a respective selection operation of the object in the 3D point cloud.

Abstract: A computer implemented method for designing a 3D object in a scene including obtaining a 3D grid of cubes in the scene, said 3D grid of cubes controlling a subdivision surface, said subdivision surface modeling the 3D object, receiving a user stroke of a curve on at least one external face of at least a cube of the grid of cubes, determining a first set of cubes which intersect with the curve on the external face, a second set of cubes which is adjacent to the first set of cubes, perpendicularly to the external face, and a third set of cubes, called intersected cubes, comprising the first set of cubes and the second set of cubes, and for each intersected cube, deforming the intersected cube by moving at least one vertex of said intersected cube so as to fit the curve thereby deforming the subdivision surface.

Abstract: The present invention is a system, comprising: isolate at least one roof truss from a building model; process a first set of data associated with the characteristics of the roof truss members, wherein the characteristics are related to the profile and length; process a second set of data associated with the assembly of the roof truss; create an assembly process of the roof truss, wherein an assembly and disassembly of the roof truss is performed to determine at least one assembly process based on the order of connecting the roof truss members from a construction perspective; formulate a package of the roof truss, wherein the package is organized based on assembly process; adjust the orientation of the members within the package based on the shipping vessel limitations and the other bundles of the model which are contained within the shipping vessel; and generate a graphical representation of the package.

Abstract: The present disclosure is directed to a software tool that facilitates coordination between various parties that are involved in the process of rectifying a problem identified in a combined three-dimensional model file. In one implementation, the software tool may cause a computing device to (a) receive an indication requesting creation of a coordination issue that relates to a portion of a rendered three-dimensional view of a construction project, (b) in response to the receipt of the indication, create a data set defining the coordination issue, the data set including (i) a representation of the portion of the rendered three-dimensional view, and (ii) data indicating an assignee of the coordination issue, and (c) cause an indication of the coordination issue to be presented to a client station associated with the assignee.

Abstract: Methods and apparatus for display of digital content associated with a location of a Node in a wireless communication area. A user interface is generated with a digital representation of content associated with a physical tag or other Node. The user interface includes interactive portions based upon positions in the wireless communication area. The interactive portions may be activated to display the digital content associated with the physical tag in a user interface.

Abstract: Simulation methods and systems are described for topology and shape optimization of a geometrical representation of a physical object being modeled. An initial geometry of the physical object is represented. Equation data representing physical properties for portions of an initial geometry are defined. User inputs are received through a GUI for optimization settings and user selections of portions of the geometry being optimized. For topology optimization, various material properties and solver settings may be defined. For shape optimization, various solver settings and optimization objective may be defined. A discretized model is generated of the physical object. Solutions of the topology or shape optimization are generated and updated values are determined. Iterative operations are then performed for solving the topology or shape optimization objective expressions. The solutions are stored and graphical representations of the solutions may be generated.

Abstract: A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform certain acts. The acts can include obtaining a route for delivering one or more orders in a tri-temp trailer from a distribution center to physical stores in a sequence of stops. The route can have an associated assignment of stack groups comprising stacks of pallets. The acts also can include determining a load design for the stacks in the tri-temp trailer based on the sequence of the stops in the route, such that unloading each of the stacks is minimized when the tri-temp trailer delivers the orders to the physical stores. The acts additionally can include outputting the load design to cause the stacks to be loaded in the tri-temp trailer according to the load design for delivery to the physical stores in the sequence of stops. The load design can specify a respective floor spot assignment for each of the stacks.

Abstract: A three dimensional (3D) backshell component is flattened to a two dimensional (2D) representation while maintaining a connected wiring component in 3D. Sketch segments for a curved 3D backshell connected first route segment within the backshell housing are stored. A first tangent is computed for a first entry point at a first end point of the connected first route segment, and a flattened route is calculated for route segments unconnected to the backshell. A flattened route position and a second tangent are calculated for a second route segment connected with the first route segment at a second entry point corresponding to the first entry point. The first entry point and the second entry point are aligned, and the first tangent and the second tangent are aligned, and the flattened unconnected route segment aligned with the 3D backshell component is displayed.

Abstract: The disclosure notably relates to a computer-implemented method for displaying a simulation. The method includes computing a full simulation. The full simulation includes states. The method further includes computing a reduced model of the computed full simulation. The reduced model includes a basis with elements. Each state of the full simulation is represented by a respective linear combination of basis elements. The method further includes displaying, for at least one state of the full simulation, a part of the respective linear combination. This constitutes an improved method for displaying a simulation.

Abstract: A system and method is disclosed for predicting and comparing wear scenarios in a rail system. The method can include generating and running a contact model of the interaction between a rail and a train car to produce a simulated loading on the rail for a predetermined time period; generating and running a wear model based on the material properties and/or friction modifier properties of the rail using the simulated loading to produce a simulated wear profile of the rail for the predetermined time period; obtaining a grinding profile to be performed on the rail during the predetermined time period; and generating an updated rail profile by modifying the rail profile by the simulated wear profile and the grinding profile. The method can predict and compare crack growth over time, and provide a financial model and comparison of costs and benefits for different maintenance protocols for the rail system.

Abstract: Systems and methods for using parametric modeling to design scenarios for autonomous vehicle simulation are provided. In particular, a computing system can obtain data identifying a plurality of parameters, each parameter associated with a particular scenario component. The computing system can determine values associated with a first set of parameters in the plurality of parameters. The computing system can determine one or more parameter relationships, such that values associated with a second set of parameters in the plurality of parameters are determined, at least in part, based on the values associated with the first set of parameters. The computing system can initiate a simulation of a scenario based, at least in part, on the values associated with the first set of the parameters and the one or more parameter relationships. The computing system can determine whether the simulated autonomous vehicle has successfully completed the scenario.

Abstract: A computing device is described that obtains a representation of a target ear canal of a user. Using a machine-learned model that has been trained based at least in part on representations of previously fabricated ear-wearable devices, the computing device generates a representation of an ear-wearable device for the target ear canal.

Abstract: A method and an apparatus for setting an evacuation exit under a crowd emergency evacuation scenario are provided. A simulation model is constructed based on a social force model, and a behavior parameter of an evacuating crowd under the crowd emergency evacuation scenario is obtained by a numerical approximation method with actual trajectory data of individuals in the evacuating crowd under a specific emergency evacuation scenario as a reference; an evacuation exit set is set, an integer programming solution for evacuation exit decision under the emergency evacuation scenario is constructed, and a total evacuation cost of the evacuation exit is obtained based on the simulation model; and reduction of the total evacuation cost is taken as an optimization target to construct an emergency evacuation optimization model based on constraint conditions, and the emergency evacuation optimization model is solved to obtain a recommended integer programming solution for evacuation exit decision.

Abstract: Conventionally, manufacturing of molded parts using composite materials has led to poor dimensional accuracy and tensile strength due to improper curing thus resulting in rejection or early/premature failure of composite part. Embodiments of the present disclosure provide simulation-based systems and methods for manufacturing/generating molded parts using reinforced composite materials. The optimized cure cycle is computed for a given component without carrying out numerous experiments. The present disclosure implements multiscale method and surrogate modeling in virtual testing for more accurate and faster manufacturing of molded parts. Process parameters for specified qualities (e.g., minimum residual stresses, minimum deformation, etc.) required for a part are determined along with least process manufacturing time. The resulting optimized time dependent cure cycle for each thermal zone of the heated mold is transferred to a master controller (e.g.

Abstract: System and method that allow utilize machine learning algorithms to move a micro-object to a desired position are described. A sensor such as a high speed camera or capacitive sensing, tracks the locations of the objects. A dynamic potential energy landscape for manipulating objects is generated by controlling each of the electrodes in an array of electrodes. One or more computing devices are used to: estimate an initial position of a micro-object using the sensor; generate a continuous representation of a dynamic model for movement of the micro-object due to electrode potentials generated by at least some of the electrodes and use automatic differentiation and Gauss quadrature rules on the dynamic model to derive optimum potentials to be generated by the electrodes to move the micro-object to the desired position; and map the calculated optimized electrode potentials to the array to activate the electrodes.

Abstract: This application relates to apparatus and methods for generating, and executing, surrogate models. In some examples, a computing device generates and evaluates correlations between input and output variables for a system to identify highly correlated input and output parameters. In addition, weights for one or more of the parameters may be determined. The computing device identifies a mathematical relationship between input and output variables to generate a physics model. The computing device may also identify other features not captured by the physical relationship that are highly correlated to each other, and generates a feature model that is based on the highly correlated features. The computing device may optimize the feature model based on a culling process that reduces the computational resources required to execute the feature model. The physics model is then combined with the feature model to generate a system output model that can simulate the system.

Abstract: Computer implemented methods include providing a numerical simulation solid mesh comprising a plurality of solid nodes and a numerical simulation fluid mesh comprising a plurality of fluid nodes, and performing a numerical simulation with the solid mesh and fluid mesh including: determining a fluid-solid interaction between the fluid mesh and the solid mesh at a time step by (i) immersing boundaries of the solid and fluid meshes, (ii) computing current corrected fluid and solid velocities using an interaction parameter applied to an intermediate fluid-solid relative velocity, and (iii) computing a fluid-solid interaction force based on the current corrected fluid and solid velocities.

Abstract: Provided are a method and system for visualization of a 3D electronic device. The method includes: a structural and/or electrical characteristic of the electronic device is simulated to obtain first visualization data, the electronic device including a semiconductor device and/or an integrated circuit and the first visualization data including 3D grid position information of the electronic device and/or a physical quantity at a grid point; the 3D grid position information and/or physical quantity at the grid point in the first visualization data are/is converted into second visualization data suitable for virtual 3D displaying according to a data type; and the second visualization data is rendered in a virtual space to display a structure and/or physical quantity of the electronic device in the virtual space.

Abstract: A device verification method, a UVM verification platform, an electronic apparatus, and a storage medium are provided. The method includes: determining a transaction class corresponding to a device under test, and instantiating a first interface in a callback function; sending input data to the device under test based on a bus protocol, and sequentially adding the input data to an array of the first interface according to addresses of the input data; instantiating a second interface in a monitor device, and sequentially adding output data to an array of the second interface according to addresses of the output data; and comparing the input data and the output data that have same addresses in the array of the first interface and the array of the second interface, and outputting a verification result of the device under test according to a comparison result.

Abstract: For each circuit element in a pair of launch and capture paths, a parameter value of the circuit element may be modified by a variation amount that is assigned to a class of circuit elements to which the circuit element belongs. Next, a timing slack may be computed for the pair of launch and capture paths.

Abstract: A method of generating an IC layout diagram includes abutting a first row of cells with a second row of cells along a border, the first row including first and second active sheets, the second row including third and fourth active sheets, the active sheets extending along a row direction and having width values. The active sheets are overlapped with first through fourth back-side via regions, the first active sheet width value is greater than the third active sheet width value, a first back-side via region width values is greater than a third back-side via region width value, and a value of a distance from the first active sheet to the border is less than a minimum spacing rule for metal-like defined regions. At least one of abutting the first row with the second row or overlapping the active sheets with the back-side via regions is performed by a processor.

Abstract: A method is provided and includes several operations: forming a first group of macros in a first region, wherein the first group of macros are aligned with a first boundary of a channel that is coupled thereto through pins of the first group of macros; forming a second group of macros in the first region to align with a second boundary of the channel that is coupled thereto through pins of the second group of macros, wherein the first and second groups of macros are coupled to a first register; and forming a third group of macros in a second region different from the first region. A first macro and a second macro that are in the third group of macros are aligned with the first and second boundaries respectively. The third group of macros are coupled to a second register different from the first register.

Abstract: Embodiments include herein are directed towards a method for electronic circuit design. Embodiments may include receiving a selection of an instance associated with an electronic design at an electronic design schematic displayed on a graphical user interface. Embodiments may also include selecting a corresponding instance within an electronic design layout displayed on a graphical user interface. Embodiments may further include receiving a selection of a source topology and routing at the electronic design layout displayed on the graphical user interface, based upon at least in part, the source topology.

Abstract: An IC test engine generates a plurality of two-cycle delay test patterns that target a first set of multicycle faults and/or defects of a fabricated IC chip based on an IC design. Each two-cycle delay test pattern includes a scan-in shift window operating at a test clock frequency, and a capture window with a launch cycle and a capture cycle operating at a functional clock frequency. The IC test engine fault simulates the plurality of two-cycle delay test patterns against a second set of multicycle faults and/or defects in the IC design utilizing sim-shifting, such that a state of the IC design after at least a last two shift clock cycles of a scan-in shift in window of each two-cycle delay test pattern of the plurality of two-cycle delay test patterns are fault simulated to provide two fault initialization cycles for detection of a multicycle delay fault and/or defect.

Abstract: Disclosed is a method of generating a multi-level summary of an article. The method may comprise generating, by a computing device, a low-level summary from article-matter in an article. The method may also comprise generating, by the computing device, a mid-level summary based on the low-level summary and the article-matter. The method may also comprise generating, by the computing device, an upper-level summary based on the mid-level summary, the low-level summary, and the article-matter.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media that merge separate digital point text objects into a single merged digital text object while preserving the properties and original visual appearance associated with the digital text included therein. For example, the disclosed systems can determine point text character properties associated with the separate digital point text objects (e.g., rotations, baseline shifts, etc.). The disclosed systems can merge the separate digital point text objects into a single merged digital point text object and modify associated font character properties to reflect the determined point text character properties. Further, the disclosed systems can generate an area text object based on the merged digital point text object where the area text object includes the digital text and the font character properties.

Abstract: Systems and methods for generating a natural-language statement for a healthcare record are provided. A graphical user interface comprising a canvas region and a menu region is displayed, wherein menu region comprises a first set of one or more interactive graphical user interface menu objects configured to receive user inputs corresponding to medical information. Data is received representing a first user input comprising user interaction with one or more of the first set of menu objects, the first input indicating medical information for a patient. In accordance with the first user input, a natural-language statement is generated based on the medical information indicated by the first user input, and display of the canvas region is updated to display the natural-language statement.

Abstract: Described herein are methods and systems to generate and display interactive graphical user interfaces that receive data inputs from users while emulating a chat session. A server retrieves a set of prompts associated with an electronic form; displays, within a webpage, a first element corresponding to a prompt and a second element having at least one input element configured to receive an input from a user interacting with the webpage. When the user interacts with an input element, the server relocates the first element and dynamically revises the second element by relocating the second element and displaying inputted values received from the user interacting with the at least one input element of the second element. The server then displays a third element comprising next prompts of the electronic form.

Abstract: A domain-specific language interpreter and live updating visual interface for rapid interactive exploration, filtering, and analysis of a dynamic data set. It includes a multi-line editor that allows a user to enter and edit input on any line at any time, and a grid view display. As the user enters an expression in the multi-line editor, it continually parses and executes the expression with respect to the domain-specific language, recognizing data tags and operations. Each data tag is associated with values for identifiers of the data set, and each operation can be applied to the values. The grid view display updates with a live display of identifiers and result values for the input expression according to the current contents of the multi-line editor user input interface.

Abstract: Aspects of the disclosure relate to joining data tables. A computing platform may input two or more tables into a statistical join function, which may initiate execution of the statistical join function, and where executing the statistical join function comprises applying one or more of: an end condition function, a partition tables function, or an outer join function to generate a new table that includes information from the two or more tables. The computing platform may send, to a user device, the new table and one or more commands directing the user device to display the new table, which may cause the user device to display the new table.

Abstract: Systems, methods, and software are disclosed herein that allow multiple values to be stored and referenced individually in a spreadsheet or table cell. In an implementation, an application identifies a format of a cell in a spreadsheet with a format having multiple data points. The application receives multiple values corresponding to the multiple data points and stores each value of the multiple values in the cell as a separately referenceable value with respect to each other value of the multiple values. In some implementations, formulas in other cells or sub-routines can reference an individual one of the values in the cell.

Abstract: It is possible to achieve high analysis accuracy on the whole irrespective of a coverage ratio of a morpheme dictionary. A morpheme prediction section 38 outputs a morpheme label string as a morphological analysis result of input text by inputting a resultant vector generated for each character by a vector combination section 36 into a series predictor and sequentially performing calculation. At this point, there is used the series predictor which is subjected to learning in advance such that a morpheme label string obtained by inputting a resultant vector in which a character vector and a morpheme information vector are combined and which is generated without being combined with the morpheme information vector with a specific probability into the series predictor and sequentially performing the calculation coincides with a morpheme label string of learning data.

Abstract: Systems and methods for natural language processing are described. One or more embodiments of the present disclosure receive a document comprising a plurality of words organized into a plurality of sentences, the words comprising an event trigger word and an argument candidate word, generate word representation vectors for the words, generate a plurality of document structures including a semantic structure for the document based on the word representation vectors, a syntax structure representing dependency relationships between the words, and a discourse structure representing discourse information of the document based on the plurality of sentences, generate a relationship representation vector based on the document structures, and predict a relationship between the event trigger word and the argument candidate word based on the relationship representation vector.

Abstract: From metadata of a corpus of natural language text documents, a relativity matrix is constructed, a row-column intersection in the relativity matrix corresponding to a relationship between two instances of a type of metadata. An encoder model is trained, generating a trained encoder model, to compute an embedding corresponding to a token of a natural language text document within the corpus and the relativity matrix, the encoder model comprising a first encoder layer, the first encoder layer comprising a token embedding portion, a relativity embedding portion, a token self-attention portion, a metadata self-attention portion, and a fusion portion, the training comprising adjusting a set of parameters of the encoder model.

Abstract: Disclosed herein are system, method, and computer program product embodiments for utilizing non-RAM memory to implement machine learning configured with a meta-learning training set (small dataset), to create a common-sense predictive language model, thus boosting the performance for downstream tasks. An embodiment operates by receiving a base sentence and perturbation sentences as an input and tokenizing the input to generate a sequence of tokens. Tokens of the semantic perturbation sentences are embedded with tokens of the base sentence as contextually similar tokens pairs to generate training data and classified to capture relationships of the base sentence and the perturbation sentences to generate a classification, which is used to train a language model.

Abstract: Computer implemented methods and systems are provided for generating diverse key phrases while maintaining competitive output quality. A system for training a sequence to sequence (S2S) machine learning model is proposed where neural unlikelihood objective approaches are used at (1) a target token level to discourage the generation of repeating tokens, and (2) a copy token level to avoid copying repetitive tokens from the source text. K-step ahead token prediction approaches are also proposed as an additional mechanism to augment the approach to further enhance the overall diversity of key phrase outputs.